OFFSETS:
Listed in the order of the images. The images that I chose myself are included.
cathedral
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
2 5
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
3 12
church
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
4 25
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-4 58
clothes
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-18 25
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-38 115
emir
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
24 49
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
40 107
harvesters
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
17 60
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
14 124
icon
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
17 42
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
23 90
lady
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
9 56
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
13 120
melons
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
10 80
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
13 177
monastery
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
2 -3
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
2 3
onion_church
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
26 51
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
35 107
sculpture
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-11 33
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-27 140
self_portrait
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
29 78
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
37 175
specimens
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-38 42
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-82 108
three_generations
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
12 54
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
9 111
tobolsk
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
2 3
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
3 6
train
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
2 42
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
29 85
vase
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-2 24
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-2 113
EXTRA CREDIT:
I cropped the edges and calculated the magnitude of the gradient (from 2 convolutions) for edge detection purposes. Then, I used NCC. This worked better than just NCC.
These were the results with just NCC (first four images):
cathedral: exact same
church: exact same
clothes: off by one (116 instead of 115; no discernible difference)
emir: big error:
aligning green and blue pics
horizontal and vertical displacement needed to match green or red to blue:
24 49
aligning red and blue pics
horizontal and vertical displacement needed to match green or red to blue:
-203 95 instead of 40 107
For pics, see first 3 pics. You may have to scroll horizontally. Left pic is original. Middle pic is with edge detection with gradient magnitudes and NCC. Right pic is with just NCC.
If you want to see the shifted green image on top of the blue image and the shifted red image on top of the blue image, see README.txt.
I calibrated an adjusting depth that depends on the size of the image so that the program can detect shifts as large as possible without picking the wrong peak early on in the pyramid algorithm and being completely off.
